Religious Concepts are Probably

Epiphenomena: A Reply to
Pyysiäinen, Boyer, and Barrett
J ESSE M. B ERING ¤

It is to the commentators’ great credit that they have managed to cull

the conceptual imperfections of my article “The Natural Foundations of
Afterlife Beliefs” (Bering 2002a) into a fairly coherent discussion of the
“innateness” of representations of dead agents’ minds. In response to their
insightful remarks, I will do my best to clarify my position on precisely this
topic of the etiology of religious concepts.1 Unlike Barrett, who stated that
“the chicken-and-egg problem of whether intuitive ideas Ž ll in explicit ones
or the other way around may be moot” (p. *), I cannot think of a more
important, central question for researchers to be concerned with.
Still, because experimental research on the cognitive bases of religion
has only recently begun in earnest (e.g., Barrett 1998; Barrett & Keil 1996;
Barrett, Richert & Driesenga 2001; Bering 2002; Bering & Bjorklund 2003;
Barrett & Nyhof 2001; Boyer & Ramble 2001; Evans 2001; Kelemen
1999; Norenzayan & Atran, in press; Walker 1992; Woolley & Phelps
2001), to some extent I feel it premature to forcibly argue the position

¤ Jesse M. Bering, Department of Psychology, University of Arkansas, Fayetteville, AR

72701; e-mail: jbering@uark.edu

1 Although my response will focus on theoretical implications, it is important to point

out an interpretive error in Pyysiäinen’s commentary. Pyysiäinen comments that, “The

subjects were only asked whether the dead person of the stimulus story ‘still’ was angry,
etc., or whether (s)he was ‘now’ happy. Answering these questions does not imply any
stance on the question of a complete cessation of emotions: : :” In fact, the methodological
design (which included critical follow-up questions) was especially sensitive to just this issue,
and the discontinuity measure re ected cessation of ability to experience such states, not the
transient absence of particular states after death, e.g., see “Coding of Interview Sessions”
(pp. 279-280).

c Koninklijke Brill NV, Leiden, 2003

° Journal of Cognition and Culture 3.3
 RELIG IOUS CONCEPTS ARE PROBABLY EPIPHE NOMENA 245

to which I am most inclined, which is that the human mind possesses

architectural constraints (cf. Elman et al. 1996) that act as probabilistic
tributaries toward implicit forms of causal reasoning that are functionally
indistinguishable from explicit, declarative religious beliefs. Furthermore,
although I view human cognitive systems as being highly receptive to
the same types of salient, counterintuitive concepts that Pyysiäinen (2001),
Boyer (1994, 2001), and, to a lesser extent, Barrett (2000; Barrett & Nyhof
2001), claim underlay all categories of religion, I remain unconvinced that
the ontogenetic expression of implicit religious beliefs turns on children’s
conceptual slots being Ž lled by such counterintuitive representations.

On the origins of ghosts and gods

Rather, it seems equally, if not more (based on recent data, Barrett

et al. 2001; Bering 2002a; Bering & Bjorklund 2003; Kelemen 1999),
plausible that the generativity of religious concepts is nothing more than
an epiphenomenal process that maps descriptive, memorable ontological
properties onto already existing causal inferences that are spontaneously
generated by individual minds (for related discussion, see Kelemen, in
press). In other words, default inferences that are typically associated with
religious thinking (e.g., belief in the continuity of personal consciousness
after death; belief in an abstract intentional agency as the arbiter of
life events, and the creator of species and natural inanimates) are not
activated by culturally transmitted religious concepts, but instead give rise
to religious concepts themselves. In turn, these concepts, which are highly
variable in nature but which possess characteristics that likely conform to
the socioecological demands of the cultures harboring them (see Reynolds
& Tanner 1995), do not play any causal role in generating general
patterns of belief, but only allow declarative access to intuitive patterns
of reasoning about typically religious matters. That is, it is unclear how
culturally acquired religious concepts can actually endow individuals with
the cognitive incentive to, for instance, envision personal consciousness as
surviving death, or to envision life events (which are the “actions” of the
gods), as being purposeful or meaningful.
SpeciŽ c cultural concepts, made salient by their minimal ontological
deviance from mundane concepts, might indeed trip-wire default notions
of psychological, biological, or physical properties of agents, as cultural
246 JESSE M. BERING

epidemiologists such as Pyysiäinen and Boyer hold. This can account

for the “theologically correct” and “theologically incorrect” Ž ndings of
Barrett and Keil (1996; also Barrett 1998), who asked their subjects about
God, a culturally speciŽ c and socially acquired, explicit supernatural agent.
Likewise, the belief that there exist such things as ghosts, who are invisible
but are human nonetheless, may well occur by way of exposure to such
an explicit concept during development, and the concept might indeed
arouse an arsenal of implicit inferences about ghosts’ minds. Pyysiäinen
may be entirely right in saying that it is misleading to claim that the
cultural concept of ghost (which is but an unreŽ ned category of dead
agent), and the attributions made toward such supernatural agents, is the
same as personal beliefs about speciŽ c, recently deceased agents and the
attributions made thereof. “Ghosts are different from deceased persons in
that they come back on earth to haunt on us. Entertaining this kind of
idea must require different cognitive operations than the representation of
mere dead agents” (p. *).
Then again, what is a ghost but an invisible dead person with a
mind? My results do not resolve the question of whether belief in “ghosts,”
per se, requires gathering information about such agents through cultural
mechanisms, but they do strongly suggest that reasoning about dead agents’
minds is only superŽ cially in uenced by explicit religious beliefs (presumed
to be an artifact of social learning) about what becomes of the mind at
death. People draw inferences about dead agents’ psychological states on
the basis of their own capacity to simulate the absence of these states;
some of these inferences are easier to make than others – the belief that
a dead agent can want something encounters substantially less resistance
than the belief that a dead agent can taste something. Despite their explicit
beliefs (e.g., whether an individual classiŽ ed himself as an extinctivist or
immortalist), this general pattern held constant.

Children’s beliefs about the psychological states of dead agents

Boyer writes that these data provide “some initial evidence for the fact
that some mental properties are more ‘sticky’ than others” (p. *), but if he
means by this that these state categories within dead agent concepts are
more easily acquired, this runs counter to my position. What I am claiming,
in contrast, is that some state categories within dead agent concepts are
 RELIG IOUS CONCEPTS ARE PROBABLY EPIPHE NOMENA 247

more easily lost than others. That is, individuals’ regular, everyday cognitive
proŽ les makes them immediate “immortalists,” wherein all mental state
properties are initially envisioned as continuing after death, but that (a)
some state categories (e.g., perceptual and psychobiological states) are
prone to “falling off” of intuitive concepts of dead agents’ minds when
they come into con ict with biological knowledge and explicit concepts
about death, whereas; (b) other states (e.g., desire, emotional and epistemic
states) are more resistant to discontinuity reasoning because their absence
cannot be simulated.
Support for this position comes in the way of empirical data assessing
children’s beliefs about the psychological continuity of the same state
categories measured for adults in the target article (Bering & Bjorklund
2003; Bering, Hernández-Blasi & Bjorklund 2003). Bering and Bjorklund
(2003) asked kindergartners (M D 5 years 3 months) and late elementary
aged children (M D 11 years 8 months) about the psychological functioning
and biological imperatives of an anthropomorphized mouse that was
preyed upon and killed by an alligator (see also H. Barrett 1999).2 Although
kindergartners were nearly at ceiling for discontinuity responses on the
biological imperatives questions (e.g., “Does the mouse’s brain still work?”),
just as the older children were, they were signiŽ cantly less likely than
the older children to reason that the dead mouse’s psychological states
ceased at death. Also, unlike the older children, who demonstrated the
same pattern of discontinuity responses as the adults in the target article
(i.e., psychobiological D perceptual > desire D emotional D epistemic),
younger children failed to distinguish between any of the state categories
(e.g., they were just as likely to say that psychobiological states continued
as they were emotional states). Even questions from state categories that
were conceptually yoked to the biological imperatives questions did not
encourage discontinuity responses from the kindergartners. For example,
although nearly all children at this age understood that the dead mouse
did not need to eat food once it was dead, they believed that the dead
mouse continued to be hungry.

2 Adult participants were also included in Bering and Bjorklund (2003), and were tested
with the puppets. Findings with the adults were similar to the oldest children, and replicated
the pattern of Ž ndings reported by Bering (2002a).
248 JESSE M. BERING

This suggests that children have naïve theories about psychological

functioning after death that initially include all psychological states and
which are gradually narrowed to include a more restrictive range of func-
tioning. Similarly, Barrett, Richert, and Driesenga’s (2001) data seem to
show that, due to their immature theory of mind, preschoolers are cog-
nitively biased toward the representation of a generic omniscient agency.
However, in response to children’s concept acquisition throughout their
development, and in combination with a burgeoning theory of mind, this
concept of agency branches out into new agent Ž gures with diversiŽ ed
mental properties. Most of these Ž gures (e.g., “Mom,” “monkey,” and
“kitty”) are endowed with a more restricted range of epistemic abilities,
but culture also provides children with an explicit supernatural agent con-
cept (i.e., God) that is capable of mapping onto the initially omniscient
intuitive concept of agency held by preschoolers. For the “afterlife” data,
this conceptual narrowing comes in response to explicit religious and bi-
ological concepts about death; but again, drawing inferences about dead
agents’ intentionality seems to be orthogonal to such explicit concepts (i.e.,
they are not required to activate inferential systems).
First, if making inferences about the psychological states of dead agents
required the possession of a counterintuitive cultural concept of “spirit”
or “ghost” or “soul,” one might expect older children to be more likely
than younger children to believe psychological functioning continued after
death, because they have had more secular and religious indoctrination.
In fact, in a replication of the study by Bering, Hernández-Blasi, and
Bjorklund (2003) that included Spanish children enrolled in religious
schools, preschoolers and kindergartners were also more likely than older
children to report the psychological continuity of dead agents. Second,
it is striking how few children used eschatological terms (e.g., “heaven,”
“ghost,” “spirit,” “God” and so on) in answering the experimenter’s
questions. Although this does not rule out the possibility that such explicit
concepts played a role in children’s answers, it does suggest that children
were predominantly relying on inferential means that were unrelated to
what they had been taught about dead agents.
 RELIG IOUS CONCEPTS ARE PROBABLY EPIPHE NOMENA 249

Hyperactive agency detectors are environment-bound

As Boyer rightly points out, it is incorrect to speak of culture as being a

separate entity that contains and transmits concepts to individuals, wherein
children’s brains might be viewed as passive receptacles of especially
potent ideas – in this context, ones of religious varieties. Evolved cognitive
mechanisms are not only receptive to but are also reactive to social
environments; they are adaptively organized in response to learning, and
constrain the broad parameters of speciŽ c cultural concepts to promote
ready and rapid inferences about the way the world works (see Bjorklund &
Pellegrini 2002). Particular ideas circulating in a given culture may indeed
primarily be ‘captured by’ and ‘adhere to’ evolved cognitive templates, but
this does not address the question of where the inferential mechanisms
supporting supernatural concepts (which are the real cognitive engines
behind religious beliefs and behaviors) come from. Rather, it is more
or less assumed that these inferential mechanisms are indistinguishable
from those responsible for people’s reasoning about intentional agents (e.g.,
orcas and lemurs and physicians) and events (e.g., people listening to our
conversations and responding to our requests through their behaviors) in
the everyday world (Atran 2002; Barrett 2000; Boyer 1994, 2001; Guthrie
1993). Guthrie (1993) and Barrett (2000), for instance, have argued that
humans have a sort of hyperactive agency detection which primes them for
perceiving intentional agency in the natural environment. This capacity is
selectively advantageous, because, as Guthrie (1993) notes, “It is better
for a hiker to mistake a boulder for a bear, than to mistake a bear
for a boulder” (p. 6). When it comes to religion, these authors claim
environmental information containing agency relevant stimuli (e.g., a stick
breaking in the forest), in the perceptual absence of natural agency, will
encourage the representation of supernatural agent concepts (e.g., a forest
spirit).
But when it comes to how people actually make religious attributions,
this explanation for the mechanisms behind spontaneous agency attribu-
tions is insufŽ cient, for the following reasons. Although environmental
events associated with agency, such as movement and noise, inarguably
promote attributions of intentionality, causal reasoning that taps into reli-
gious beliefs is hardly limited to these objective (e.g., environment-bound)
event categories. Rather, the bulk of people’s reasoning about supernat-
250 JESSE M. BERING

ural agency concerns events that bear narratively on people’s lives, such as
subjectively traumatic experiences (McAdams, Josselson & Lieblich 2001).
These experiences are conceptually distinct from the events that typically
co-occur with intentional behaviors (e.g., walking in the forest causes sticks
on the ground to break), because they are abstract (e.g., “losing a child”),
subjective (e.g., winning the lottery might be seen as a blessing for one
person, and a curse for another), and are not always environment-bound
(e.g., we might perceive our spouse to be angry, when in fact she is not).
For example, Gilbert, Brown, Pinel & Wilson (2000) showed that people
frequently “subjectively optimize” negative outcomes that they have no
control over, oftentimes mistaking the workings of their own psycholog-
ical immune systems (which are designed to guard against anxiety and
loss of self-esteem caused by environmental assaults) with the benevolent
intentions of an external agency (e.g., God).

Wherein lies meaning, and the representational self, in

cognitive models of religion?

What each of the commentators have failed to demonstrate in their

previous work is how events occurring “out there” in the objective
environment are indistinguishable from events occurring “in here” in the
subjective environment. Reasoning that a stick breaking in the middle of
a dark forest is caused by a supernatural agent is critically different from
reasoning that one’s miscarriage is caused by a supernatural agent; thinking
that an aspen tree that is shivering human-like in the wind is watching us
is different from thinking that the tree is an omen carrying a message for
us. Only the second case in each comparison requires social cognitive skills
capable of disambiguating communicative messages from an intentional
agency – or, in other words, deriving existential meaning from perceptual
experiences (Bering 2002b). This form of meaning has been altogether
absent in contemporary cognitive models of religion, which is surprising
considering that it is likely at the heart of individual people’s everyday
experiences with religion. Most theorists in this area, rightly eschewing the
muddled framework of earlier relativistic accounts that viewed religion as
private and subjective, have adopted a stance that views human minds as
astretegic and non-social in nature. Although scholars such as Boyer (2001)
and Atran (2002) do note that human minds become strategic in religious
 RELIG IOUS CONCEPTS ARE PROBABLY EPIPHE NOMENA 251

matters when they are faced with speciŽ c supernatural agents that are said
to have “full-access” to their thoughts and behaviors, neither evolutionary
theorist spends time articulating the cognitive mechanisms that allow (and
impel) people to make sense of random events as they relate to the self, discuss
under what subjective circumstances events are perceived as being about
the supernatural agent’s intentional states, or deŽ ne the event parameters
that are likely to solicit meaning-making from individual minds.
In general, current models lean heavily on cultural mechanisms of
concept transmission, viewing individual minds as “input-output” meme-
processing and replicating machines. Contrary to Pyysiäinen’s claim that
cultural epidemiologists “emphasize the powers of individual minds more
than what Bering thinks” (p. *), there is, in fact, a noticeable absence in
the role of the self in the development of reasoning about religious matters.
Concepts are indeed portrayed as coming from outside the head and
“exploiting” or “parasitizing” templates meant for processing information
in core ontological domains, and these processes are seen as constituting
the “natural foundations of religion.” To me, the Ž ndings borne of the
simulation constraint hypothesis on children and adult’s intuitive reasoning
about death serve as a clear departure from the dominant “counterintuitive
concept” theoretical skein, in that it is one of the Ž rst examples to show the
active, generative processes of individual minds in forming proto-religious
beliefs. In some cases, these beliefs occur in spite of explicit religious
concepts.
But the real distinction between my own perspective and those of the
commentators (at least Pyysiäinen and Boyer) is that only the latter be-
lieve that the tendency to make inferences about supernatural agency is
dependent upon the presence of explicit concepts that have been intro-
duced through social channels. Boyer (2001, p. 40), for instance, writes
that “people have religious notions and beliefs because they acquired them
from other people : : : people get their religion from other members of their
social group.” In contrast, my own view is that the same automatic infer-
ences driving the real-time representation of supernatural agents occurs
independent of such explicit religious concepts.
There is increasing evidence that such content-free inferential processes arise
through individual cognitive systems (which includes self-awareness) being
confronted with broad categories of experiential invariants that character-
252 JESSE M. BERING

ize early human development (Spelke & Newport 1998; Wynn 1992). The
acquisition of explicit concepts through cultural means of transmission ap-
pears to be critical for “Ž lling in” these general inferential processes with
content enriched information about agency, but the inferential processes
themselves are neither enabled by nor activated by such concepts. Instead,
explicit religious concepts might be epiphenomena that shadow the op-
erations of intuitive patterns of reasoning. My current view is that these
patterns of reasoning are functionally indistinguishable from those that oc-
cur under the cultural and academic label of religion.

REFERENCES

A TRAN , S.
2002 In gods we trust: The evolutionary landscape of religion. Oxford: Oxford University
Press.
B AR RETT , J.L. & K EIL , F.C.
1996 Conceptualizing a non-natural entity: Anthropomorphism in God concepts.
Cognitive Psychology 31, 219-247.
B AR RETT , J. & N Y HOF , M.
2001 Spreading nonnatural concepts. Journal of Cognition and Culture 1, 69-100.
B AR RETT , J.L.
2000 Exploring the natural foundations of religion. Trends in Cognitive Sciences 4, 29-34.
1998 Cognitive constraints on Hindu concepts of the divine. Journal for the ScientiŽc
Study of Religion 37, 608-619.
B AR RETT , J.L., R ICH ERT , R.A. & D RIESE NG A , A.
2001 God’s beliefs versus mother’s: The development of nonhuman agent concepts.
Child Development 72, 50-65.
B ERING , J.M.
2002a Intuitive conceptions of dead agents’ minds: The natural foundations of afterlife
beliefs as phenomenological boundary. Journal of Cognition and Culture 2, 263-308.
2002b The existential theory of mind. Review of General Psychology 6, 3-24.
B ERING , J.M. & B JO RKLU N D , D.F.
2003 Simulation constraints and the natural emergence of afterlife reasoning as a
developmental regularity. Manuscript in submission.
B ERING , J.M., H ERNÁN D EZ -B LAS I , C. & B JO RKLU N D , D.F.
April, 2003 The role of metarepresentational-simulationist models in the development of death
reasoning. Poster presented at the biennial meeting of the Society for Research in
Child Development, Tampa, FL.
B JO RKLU N D , D.F. & P ELLE GRIN I , A.D.
2002 The origins of human nature: Evolutionary developmental psychology. Washington, DC:
American Psychological Association.
 RELIG IOUS CONCEPTS ARE PROBABLY EPIPHE NOMENA 253

B O YE R , P. & R A MB LE , C.
2001 Cognitive templates for religious concepts: Cross-cultural evidence for recall of
counter-intuitive representations. Cognitive Science 25, 535-564.
B O YE R , P.
1994 The naturalness of religious ideas: A cognitive theory of religion. Berkeley: University of
California Press.
2001 Religion explained: The evolutionary origins of religious thought. New York: Basic Books.
E LMA N , J.L., B A TES , E.A., J O HN SON , M.H., K AR MILOFF -S M ITH , A., P AR ISI , D. &
P L UNKE T , K.
1996 Rethinking innateness: A connectionist perspective on development. Cambridge, MA: MIT
Press.
E VAN S , M.
2001 Cognitive and contextual factors in the emergence of diverse belief systems:
Creation versus evolution. Cognitive Psychology 47, 217-266.
G ILB ERT , D.T., B RO W N , R.P., P INE L , E.C. & W ILSO N , T.D.
2000 The illusion of external agency. Journal of Personality and Social Psychology 79, 690-
700.
G U THRIE , S.
1993 Faces in the clouds: A new theory of religion. New York: Oxford University Press.
K ELEM EN , D.
(in press) Are children “intuitive theists”?: Reasoning about purpose and design in
nature. Psychological Science.
1999 Why are rocks pointy? Children’s preferences for teleological explanations of the
natural world. Developmental Psychology 35, 1440-1452.
M C A DA MS , D.P., J O SSE LSON , R. & L IEBLI CH , A.
2001 Turns in the road: Narrative studies of lives in transition. Washington, DC: American
Psychological Association.
N O REN ZAYAN , A. & A TRAN , S.
(in press) Cognitive and emotional processes in the cultural transmission of natural
and nonnatural beliefs. In M. Schaller and C. Crandall (Eds.), The psychological
foundations of culture. Hillsdale, NJ: Erlbaum.
P YY SIÄIN EN , I.
2001 How religion works: Towards a new cognitive science of religion. Cognition and Culture Book
Series, 1. Leiden: Brill.
R EYN OLD S , V. & T AN NE R , R.
1995 The socioecology of religion. New York: Oxford University Press.
S PELK E , E.S. & N EWPO RT , E.L.
1998 Nativism, empiricism, and the development of knowledge. In W. Damon (Gen.
Ed.) & R. Learner (Ed.), Handbook of child psychology: Vol. 1. Theories of theoretical
models of human development (5th ed., pp. 275-340). New York: Wiley.
W ALK ER , S.J.
1992 Supernatural beliefs, natural kinds, and conceptual structure. Memory & Cognition
20, 655-662.
254 JESSE M. BERING

W OO LLE Y , J.D. & P H ELP S , K.E.

2001 The development of children’s beliefs about prayer. Journal of Cognition and Culture
1, 139-166.
W YNN , K.
1992 Addition and subtraction by human infants. Nature 358, 749-750.